Reference Publication: Bouchelle, M., Parker, D., Anello, M., "Factors Influencing Water Heating Energy Use And Peak Demand In A Large Scale Residential Monitoring Study," Presented at: The Symposium on Improving Building Systems in Hot and Humid Climates, San Antonio, TX, May 15-17, 2000. Disclaimer: The views and opinions expressed in this article are solely those of the authors and are not intended to represent the views and opinions of the Florida Solar Energy Center. |
Factors
Influencing Water Heating Energy Use And Peak
Demand In A Large Scale Residential
Monitoring Study
Matthew
P. Bouchelle, Danny S. Parker, Michael T. Anello
Florida
Solar Energy Center (FSEC)
FSEC-CR-1671-00
Updated PDF version of the paper
Abstract
A load
research project by the Florida Power Corporation (FPC) is monitoring
204 residences in Central Florida, collecting detailed end-use load
data. The monitoring is being performed to better estimate the impact
of FPC's load control program, as well as obtain improved appliance
energy consumption indexes and load profiles. A portion of the monitoring
measures water heater energy use and demand in each home on a 15-minute
basis. The paper summarizes the various impacts identified on water
heating energy use and demand.
Hot Water Electric Demand and Consumption
The majority (153) of the water heating systems in the project were of the conventional electric resistance type. Seventeen of the monitored homes have natural gas or propane water heat and have no electric demand. These sites were eliminated from further analysis. Twenty eight (14%) of water heaters in the monitoring project have heat recovery units. There are also four operating solar water heating systems. There is also one heat pump water heater and one tank-less water heater (Site 18). Eighty percent of water heaters were located in unconditioned spaces -- primarily in garages. Eighteen percent were located inside the conditioned zone.
Table 1 summarizes the recorded winter energy use and demand against selected water heating characteristics. Demand within the table is for the hour between 7 and 8 AM on January 5th, 1999, the coldest morning when no load control was applied.
The summary statistics on hot water heating showed that occupancy has the strongest influence on variation in energy consumption. Accordingly, within the table we also normalized water heating energy use and peak demand by number of household occupants. This showed that the apparent influence of tank size on peak demand resulted from the natural association between tank capacity and household size.
Beyond
household characteristics, the water heating data revealed several
important influences that may represent opportunities for FPC to meet
its winter load control objectives.
Pipe insulation did not show up to be significant. We speculate that this may be due to the very short plumbing runs. Super insulated tanks did not show to be significant likely due to the very small sample size. The issues of HRU performance and the impact of external tank insulation is examined in greater detail in the following sections.
Figure 1 shows a histogram of the frequency distribution of measured hot water energy use within the project sample.
Table 1. Effect of Selected Characteristics on Winter
Electric Water Heating Energy Use and Demand.
Characteristic | kWh Day |
n | kW | n | kWhD/ Occupant |
n | kW/ Occupant |
n |
Type Electric
Resistance HRU Solar |
7.69 8.34** 3.11* |
154 26 4 |
0.718 0.777** 0.237* |
129 20 4 |
3.14 3.48 1.97* |
151 24 4 |
0.293 0.324 0.080* |
126 19 4 |
Occupants? =1 =2 =3 =4 >4 |
4.36* 6.52 9.48 10.22 10.37** |
25 74 25 30 23 |
0.475* 0.508 0.750 1.037 1.137** |
23 56 21 26 22 |
NA NA NA NA NA |
NA NA NA NA NA |
NA NA NA NA NA |
NA NA NA NA NA |
Hot
Water Timer? Yes No |
6.46* 7.89 |
27 156 |
0.653 0.722 |
19 134 |
3.01 3.20 |
26 152 |
0.316 0.289 |
18 130 |
Tank
Size? <40
gal =40 gal >40 gal |
5.768* 8.180 7.630 |
27 109 47 |
0.579 0.681 0.859** |
21 92 40 |
3.15 3.27 2.93 |
27 107 44 |
0.362 0.268 0.313 |
21 90 37 |
Element
Size? >4
kW <4 kW >3 kW < 3 kW |
7.99 6.62* 4.76 |
153 22 8 |
0.731 0.723 0.233 |
128 20 8 |
3.24 2.83 2.45 |
151 21 6 |
0.284 0.388 0.051 |
126 21 3 |
Conditioned
Space? Yes No |
7.99 7.62 |
30 153 |
0.524* 0.754 |
27 126 |
2.81* 3.24 |
29 149 |
0.220* 0.308 |
26 122 |
External
Insulation? Yes No |
6.32* 7.92 |
27 156 |
0.501* 0.753 |
24 129 |
3.05 3.19 |
27 151 |
0.219* 0.307 |
24 124 |
Super
Insulation? Yes No |
7.58 7.69 |
13 170 |
1.005 0.686 |
13 140 |
2.81 3.20 |
13 165 |
0.367 0.285 |
13 135 |
Pipe
Insulation? Yes No |
7.91 7.62 |
39 144 |
0.761 0.700 |
39 120 |
3.25 3.15 |
36 142 |
0.317 0.286 |
30 118 |
* Significantly lower at > 90% level ** Significantly greater at > 90% level
Seasonality of Water Heating Loads
Although water heating is not totally dominated by weather like space heating, these loads are still sensitive to temperature conditions.
The first plot (Figure 2) shows how daily average hot water energy use varied in the sample by the daily average air temperature measured in the project. Admittedly, there is considerable scatter. However, simple linear regression plotted explains 58% of the variation in the day-to-day hot water energy consumption. Moreover, including a dummy variable for week-ends does nothing for the regression. DHW use is just slightly higher on weekends and the demand profile differ, however this is not nearly as great as that of temperature.
Figure 3 shows the daily average 15-minute power DHW demand profile for the 183 sites with valid data for two days: January 5th the coldest non-load control day and March 31st, the warmest of the winter days analyzed. Average hot water energy con-sumption was 30% higher on the cold day based on variation in temperature. The graph shows that most 15-minute intervals had higher demand on the cold day.
Figure 3. Water heating load profiles on two days with
coldest and warmest conditions.
There are several reasons for this influence:
Moreover, the data series through March does not reflect the complete picture. The summer data, now being collected, should show even greater weather related impact for water heating. One study in the literature on load control does acknowledge the seasonality of LM impacts for water heating (Haeri and Gervais, 1992) and suggests that the load profiling or time temperature matrix (TTM) may be superior for assessment.
The current FPC TTMs for water heat vary by month, which captures some of the seasonal variation described. However, we found it necessary to produce hot water TTMs which respond to temperature, particularly to capture the elevated DHW demand on the most extreme winter days. This is important since the need for load control is highest on these days. For example, the average January DHW demand between 7 and 8 AM is only 0.54 kW. However, during hours when the temperature was near 32oF at 8 AM the ty-pical demand was 0.75 kW - a 39% increase in load.
Eligibility of water heating systems for load control is not affected by whether the homeowners have non-standard water heating systems. Many households have heat recovery units and several have solar water heaters or heat pumps water heaters. Thus, the total sample of all-electric water heaters are included in the time of day estimate of water heater load for computing the regression based load profiles. Sites with natural gas or propane water heat were not included. Estimates are contained in Table 2 for the period between January and July, 1999. The estimates in Table 2 have the form:
kWdhw = Aj + Bj(T)
Where:
Aj = Non-temperature responsive component of water heat demand (kW)
Bj = Temperature coefficient for DHW electric demand in hour "j" (kW/oF)
T = Outdoor ambient air temperature (oF)
Table
2. FPC Residential Monitoring Project Water Heater Hourly
Demand (kW)
Values (n = 186)
Hour | Constant (A) | |
1 | 0.225 | - 0.00146 |
2 | 0.143 | -0.00076 |
3 | 0.159 | -0.00117 |
4 | 0.153 | -0.00109 |
5 | 0.182 | -0.001185 |
6 | 0.377 | -0.002433 |
7 | 0.855 | -0.006017 |
8 | 0.977 | -0.006959 |
9 | 0.762 | -0.004058 |
10 | 0.547 | -0.001738 |
11 | 0.508 | -0.001714 |
12 | 0.533 | -0.002499 |
13 | 0.561 | -0.003222 |
14 | 0.535 | -0.003450 |
15 | 0.465 | -0.002641 |
16 | 0.432 | -0.002316 |
17 | 0.512 | -0.002744 |
18 | 0.571 | -0.002662 |
19 | 0.779 | -0.004387 |
20 | 0.743 | -0.004064 |
21 | 0.720 | -0.004771 |
22 | 0.706 | -0.004976 |
23 | 0.462 | -0.002283 |
24 | 0.340 | -0.001502 |
The profiles in Figure
5 show the described seasonality in water heater energy demand.
The water heating loads are somewhat lower than commonly supposed. Part
of this is due to the advent of low hot water using appliances and showerheads
(EPRI, 1997). Another part of the low consumption comes from occupancy;
some homes (e.g. Site 50) were unoccupied during much of the study while
others (e.g. Site 22) turned off the water heater breaker when away
from home for extended periods.
Water heating loads are greatest during the colder months. April clearly shows the shift in timing of water heating load imposed by Daylight Savings Time. The later spring and summer months show progressively lower water heating loads.
Water Heating System Type
We examined how water heating system type influenced electric demand and energy use. Some 14% of the sample (28 sites; 26 sites with valid data) had heat recovery units which scavenge heat from the air conditioning system to heat water. Four homes had operating solar water heating systems. Figures 6 and 7 suggest some interesting facets concerning the operation of these water heating systems.
Figure 7. Measured July DHW load profiles by system
type.
As expected, the average demand profile in July shows that HRU water heaters used about 30% less electricity than the electric resistance group. Demand was also lower in all hours. Secondly, solar water heating systems show even better relative performance and demand reductions during the peak hour, although the sample size is small.
The situation for winter months was completely different. First, the HRU systems used more energy and produced more electric demand for water heating in winter than their electric resistance counterparts. The demand difference between the two systems from 7 - 8 AM during January was approximately 160 Watts or about a 32% increase in utility winter coincident morning demand. Further, the difference was statistically significant at a 99% confidence level.
Daily water heating energy use was also 1.0 kWh greater in the homes with HRUs (13% greater). One explanation for this difference is that HRU owners use more hot water during winter on the mistaken belief that "hot water is free." Elevated hot water consumption associated with HRU users has been observed in another comparative project in which HRUs and electric resistance systems were metered (Merrigan, 1983). However, there are also other physical explanations for the poor winter performance:
Summer data shows the advantage expected for these systems. Here, the electric resistance water heaters use about 5 kWh per day as opposed to 3.5 kWh for the HRU systems. The demand reduction from 4 - 5 PM is only 100 Watts, however. The savings in daily water heating energy use is 1.5 kWh or approximately a 30% reduction in water heating energy.
Annually, however, the advantage of HRU systems may be marginal, both for the utility and for the consumer. Over the period from January - July, the average consumption for electric resistance water heating systems was 6.36 kWh/day as opposed to 6.23 kWh/Day for the HRU systems (suggesting annual DHW energy use of 2320 and 2270 kWh, respectively). Although water heating energy is saved during summer, this is nearly offset by increased consumption in winter. Thus, the apparent annual energy reduction for the consumer is only a few percent.
From FPC's perspective as a winter peaking utility, the reduction in demand during the summer utility coincident peak is less than the increase in the winter coincident peak and the annual reduction in hot water energy use is very small for the consumer. Although it seems likely that a number of the systems are not functioning properly, the added capital expense may be difficult to justify. From a utility load control perspective, it seems very desirable to load manage HRU sites to gain full advantage from them - particularly given their elevated winter demand.
One obvious influence on HRU performance is the selected hot water thermostat setting. Since condenser heat temperature may be no higher than 140oF, those systems with high settings may perform poorly. Unfortunately hot water set temperature was not collected in the audit, although an exit collection of this information may be useful.
There are four operating solar water heating systems in the project. Although a small sample, they showed large reductions in coincident demand as well as energy. The reduction in seven month energy use was 52% against electric resistance systems. Utility peak coincident reductions were approximately 0.35 kW in winter and 0.10 kW in summer.
Diagnostic Evaluation of HRU Performance
Given the problems identified with HRU performance, we examined each of the sites possessing these systems to determine which sites appeared to be functioning properly. This was done by plotting daily hot water energy consumption against daily air conditioning energy consumption from January - July of 1999. Generally, one should expect to see hot water electricity consumption decline as greater air conditioning provides auxiliary heat for hot water. This trend is clearly evident in Figure 8, which shows the two values plotted for the HRU at Site #10.
We found that 12 of the evaluated HRUs fell into this category of proper function. Unfortunately, there was a group of 10 households with HRUs that showed no discernable impact of increased air conditioning use lowering hot water electric consumption. An example of this problem is shown in Figure 9.
Three other HRU sites could not be classified due to little air conditioning use or vacancy. Regardless, our cursory evaluation indicates that nearly half of installed HRU systems may not be properly operating - a likely explanation for the poor level of performance observed.
Impact of DHW Element Size on Peak Demand
Down sizing of hot water tank elements is an idea which seems as if it could impact how water system peak demand. Unfortunately, the project data showed the impact is very small.
We used data for January 5th of this year (the coldest non-load managed day) and examined how the recorded water heater electric demand varied depending on the water heater element size (reliably available in the data set from the maximum recorded kW over the entire season). The lack of impact has to do with the diversity of water heating with respect to hourly demand. Simply put, so few of the water heaters are on at the same time, that although changing an element to a smaller one will reduce the demand for that single household at the time they use hot water, it will not have much effect on the overall population since hot water draws are nearly randomly distributed over the hour-long window of interest.
For the 153 non-gas sites which had valid data from the project that morning, the average water heater electric demand was 0.713 kW. The average electric water heater element size was 4.424 kW. This implies a diversity of 16% overall - most water heaters were only on a small fraction of the time. A frequency histogram shows that over 45% of water heaters were not on during that hour in spite of no load control (Figure 10). Many of these systems were likely on the hour before or after the hour examined (related to diversity of occupant showers, schedules, absence, etc.).
Figure 10. Histogram of DHW electrical demand at 7 -
8 AM on January 5, 1999.
Note that fully 45% of tanks require
no power during this hour.
To look into element size, we segmented the data into two groups: one with the element size was between 4 and 5 kW and another where the element size was between 3 and 4 kW. We then compared the hourly average demand in the two groups:
Element Size |
Avg
Element Size |
Diversified kW |
n |
4-5 kW | 4.586 | 0.7266 | 122 |
3-4 kW | 3.558 | 0.7229 | 20 |
Although the sample sizes are very different, the diversified kW is nearly identical and a statistical t-test of means showed no meaningful difference.
A second estimate utilizes a duty cycle approach with the histogram in Figure 10. Limiting element size to 3.5 kW would only impact the five water heating systems (3% of the population) whose average hourly demand was greater than that value. Applying the duty cycle method estimates an average population demand reduction of only 15 watts.
As a final check, we censured the sample to only those systems that had some power draw on the DHW circuit during the peak hour:
Element Size | Avg Element Size | Diversified kW | n |
4-5 kW | 4.596 | 1.248 | 71 |
3-4 kW | 3.558 | 1.205 | 12 |
The 40 watt difference is in the expected direction, but still shows no statistical significance (t = 0.134) with a small sample size. A non-parametric test of medians (Wilcoxon Rank Sum Test) showed that while there may be a small difference from a smaller element size, the difference is very small. The duty cycle assessment above is likely the most accurate estimate. A change in element size would affect the magnitude of the payback spike after load control when diversity is very low (most controlled systems are on). However, this would hardly be worth a program. Abatement in the DHW control release payback spike could be easily achieved through control release of three blocks of water heaters over succeeding 15-minute periods.
Hot Water Tank Wrap
Evidence that emerges from the analysis of the FPC data on water heater electric demand is that exterior tank wraps show large impacts on the measured hot water tank electrical demand, yet a much lower influence on energy use. This can be exploited to help control winter peak demand.
Theory/Laboratory Measurements
Detail measurements of hot water tank standby losses were performed in an environmental chamber by Ek at the Bonneville Power Administration (1984). He showed that electric storage tanks of the modern type have a heat loss coefficient of approximately 0.93 W/oF. When an R-11 exterior tank wrap is added, the loss coefficient drops to approximately 0.65 W/oF. With a hot water tank temperature of 130oF and a surrounding temperature of 40oF (e.g. an unconditioned garage or utility room), the average reduction in tank standby losses from an exterior tank wrap should amount to approximately 25 W.
Field Estimates
There were 26 sites within the project sample which included external tank insulation wraps. The average demand of these sites on January 5th between 7 and 8 AM when the outdoor temperature was 37oF was 0.501 kW. This compares to 0.753 kW in the sample without and external insulation wrap. The difference 0.252 kW is significant at the 90% level but is very different from the value predicted by laboratory measurement. This may be because changing the heat loss rate of the tank significantly alters diversity so elements are not immediately activated when hot water is drawn. Further, the differences still remain after controlling for household occupancy - the largest carrier of variation within water heating data. If solar water heating systems and tanks located within the conditioned space are excluded from the control sample, the estimated savings increases further (0.40 kW peak reduction). Finally, a photographic review of the hot water tank wraps in the monitoring project show that at least half of the applications are marginal (partial tank wraps, insulation missing, etc.). A utility sponsored program should be able to choose effective insulation kits (e.g. Consumer Reports, 1981) and lead to effective applications as those shown above at Site 24. Thus, hot water tank wraps look to have a large potential impact on winter peak hot water power demand for FPC if costs of installations can be made low.
The measured
reduction to annual hot water heating energy would entail some loss
of revenue. Within the winter data, the average reduction in daily
water heating energy consumption was 1.6 kWh/day. This number will
be no more than half this value for the overall year, since ambient-tank
temperature differences are much lower at other times. Regardless,
the measure would have the simple benefit of reducing customer energy
costs in a modest fashion while significantly impacting winter coincident
peak demand from non-load controlled customers.
Conclusions
The project has identified a number of influences on water heater electric demand that are not commonly described. This includes the pronounced seasonality of water heating demand load shapes as well as the time of day influence on loads. The project has also revealed that recent weather conditions have a strong influence on water heating demand beyond the normally recognized seasonal effect. A number of additional identified impacts:
References
Brecker, B.R. and Stogsdill, D. E., 1990. "A Domestic Hot Water Use Database," ASHRAE Journal, September, Atlanta, GA
Colliver, D.G., Murphy, W.E. and Taraba, J.L., 1988. "Residential Water Heater Electrical Usage and Demand Reduction using Reduced Element Sizes and Time Clock Controls," ASHRAE Transactions, Vol. 94, p. 1110-1122, Atlanta, GA.
Consumer Reports, 1981. "Insulation Kits: Wrapping Up and Saving," Consumer Reports, October, 1981, p. 590.
Ek, C.W., 1984. "Effectiveness of Conservation Measures on the Standby Energy Losses from Electric Water Heaters," ERGH-84-21, Bonneville Power Administration, Portland, OR.
EPRI, 1997. "EPRI Hot Water Consumption Study Refutes Common Assumptions," Electric Water Heating News, EPRI, Palo Alto, CA, Summer.
Merrigan, T., 1983. Residential Conservation Demonstration: Domestic Hot Water, Final Report, FSEC-CR-90-83, Florida Solar Energy Center, Cocoa, FL.
Presented at:
The Symposium on Improving Building Systems in Hot and Humid Climates
May 15-17, 2000
San Antonio, TX